Exploring Strengths and Weaknesses of Super-Resolution Attack in Deepfake Detection
Authors:
Davide Alessandro Coccomini,
Roberto Caldelli,
Fabrizio Falchi,
Claudio Gennaro,
Giuseppe Amato
Abstract:
Image manipulation is rapidly evolving, allowing the creation of credible content that can be used to bend reality. Although the results of deepfake detectors are promising, deepfakes can be made even more complicated to detect through adversarial attacks. They aim to further manipulate the image to camouflage deepfakes' artifacts or to insert signals making the image appear pristine. In this pape…
▽ More
Image manipulation is rapidly evolving, allowing the creation of credible content that can be used to bend reality. Although the results of deepfake detectors are promising, deepfakes can be made even more complicated to detect through adversarial attacks. They aim to further manipulate the image to camouflage deepfakes' artifacts or to insert signals making the image appear pristine. In this paper, we further explore the potential of super-resolution attacks based on different super-resolution techniques and with different scales that can impact the performance of deepfake detectors with more or less intensity. We also evaluated the impact of the attack on more diverse datasets discovering that the super-resolution process is effective in hiding the artifacts introduced by deepfake generation models but fails in hiding the traces contained in fully synthetic images. Finally, we propose some changes to the detectors' training process to improve their robustness to this kind of attack.
△ Less
Submitted 5 October, 2024;
originally announced October 2024.
Optical identification using physical unclonable functions
Authors:
Pantea Nadimi Goki,
Stella Civelli,
Emanuele Parente,
Roberto Caldelli,
Thomas Teferi Mulugeta,
Nicola Sambo,
Marco Secondini,
Luca PotÌ
Abstract:
In this work, the concept of optical identification (OI) based on physical unclonable functions is introduced for the first time, to our knowledge, in optical communication systems and networks. The OI assigns an optical fingerprint and the corresponding digital representation to each sub-system of the network and estimates its reliability in different measures. We highlight the large potential ap…
▽ More
In this work, the concept of optical identification (OI) based on physical unclonable functions is introduced for the first time, to our knowledge, in optical communication systems and networks. The OI assigns an optical fingerprint and the corresponding digital representation to each sub-system of the network and estimates its reliability in different measures. We highlight the large potential applications of OI as a physical layer approach for security, identification, authentication, and monitoring purposes. To identify most of the sub-systems of a network, we propose to use the Rayleigh backscattering pattern, which is an optical physical unclonable function and allows to achieve OI with a simple procedure and without additional devices. The applications of OI to fiber and path identification in a network and to the authentication of users in a quantum key distribution system are described.
△ Less
Submitted 26 July, 2023; v1 submitted 3 May, 2023;
originally announced May 2023.
